Anti-Angiogenic Effects of VEGF Stimulation on Endothelium Deficient in Phosphoinositide Ecyr Cling

Washington University School of Medicine Digital Commons@Becker

Open Access Publications

2020

Anti-angiogenic effects of VEGF stimulation on endothelium deficient in phosphoinositide ecyr cling

Amber N. Stratman

Olivia M. Farrelly

Constantinos M. Mikelis

Mayumi F. Miller

Zhiyong Wang

See next page for additional authors

Follow this and additional works at: https://digitalcommons.wustl.edu/open_access_pubs Authors Amber N. Stratman, Olivia M. Farrelly, Constantinos M. Mikelis, Mayumi F. Miller, Zhiyong Wang, Van N. Pham, Andrew E. Davis, Margaret C. Burns, Sofia A. ezP oa, Daniel Castranova, Joseph J. Yano, Tina M. Kilts, George E. Davis, J. Silvio Gutkind, and Brant M. Weinstein ARTICLE

https://doi.org/10.1038/s41467-020-14956-z OPEN Anti-angiogenic effects of VEGF stimulation on endothelium deﬁcient in phosphoinositide recycling

Amber N. Stratman 1,2, Olivia M. Farrelly1,8, Constantinos M. Mikelis3,4,8, Mayumi F. Miller1,8, Zhiyong Wang3,5,8, Van N. Pham1, Andrew E. Davis1, Margaret C. Burns1,Soﬁa A. Pezoa1, Daniel Castranova1, ✉ Joseph J. Yano1, Tina M. Kilts6, George E. Davis7, J. Silvio Gutkind 3,5 & Brant M. Weinstein 1

Anti-angiogenic therapies have generated signiﬁcant interest for their potential to combat

1234567890():,; tumor growth. However, tumor overproduction of pro-angiogenic ligands can overcome these therapies, hampering success of this approach. To circumvent this problem, we target the resynthesis of phosphoinositides consumed during intracellular transduction of pro- angiogenic signals in endothelial cells (EC), thus harnessing the tumor’s own production of excess stimulatory ligands to deplete adjacent ECs of the capacity to respond to these signals. Using zebrafish and human endothelial cells in vitro, we show ECs deficient in CDP- diacylglycerol synthase 2 are uniquely sensitive to increased vascular endothelial growth factor (VEGF) stimulation due to a reduced capacity to re-synthesize phosphoinositides, including phosphatidylinositol-(4,5)-bisphosphate (PIP2), resulting in VEGF-exacerbated defects in angiogenesis and angiogenic signaling. Using murine tumor allograft models, we show that systemic or EC specific suppression of phosphoinositide recycling results in reduced tumor growth and tumor angiogenesis. Our results suggest inhibition of phosphoinositide recycling provides a useful anti-angiogenic approach.

1 Division of Developmental Biology, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD 20892, USA. 2 Department of Cell Biology and Physiology, Washington University School of Medicine, St. Louis, MO 63110, USA. 3 Craniofacial and Skeletal Diseases Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD 20892, USA. 4 Department of Pharmaceutical Sciences, School of Pharmacy, Texas Tech University Health Sciences Center, Amarillo, TX 79106, USA. 5 Department of Pharmacology, UC San Diego Moores Cancer Center, La Jolla, CA 92093, USA. 6 Oral and Pharyngeal Cancer Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD 20892, USA. 7 Department of Molecular Pharmacology and Physiology, University of South Florida School of Medicine, Tampa, FL 33612, USA. 8These authors contributed equally: Olivia M. Farrelly, Constantinos M. Mikelis, Mayumi F. Miller, Zhiyong Wang. ✉ email: ﬂyingﬁ[email protected]

NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z

hosphoinositides are utilized as “second messengers” for a PIP2 serves as a key substrate for both phospholipase C-gamma 1 Pwide variety of different intracellular signaling pathways, in (PLCγ1) and phosphoinositide 3-kinase (PI3K) dependent sig- virtually every cell type. This includes receptor tyrosine naling downstream from multiple pro-angiogenic tyrosine kinase kinase pro-angiogenic signaling in endothelial cells (EC) through receptors, such as VEGFR2, fibroblast growth factor receptor 1 receptors like vascular endothelial growth factor (VEGF) receptor (FGFR1) and epidermal growth factor receptor (EGFR)1–8,31–34 2 (VEGFR2), the VEGFA receptor, which uses phosphatidylino- (Fig. 1b). Loss or knockdown of either one of the two CDS genes sitol-(4,5)-bisphosphate (PIP2) as a key substrate for both phos- in ECs in vitro or in zebrafish embryos in vivo results in reduced pholipase C-gamma 1 (PLCγ1) and phosphoinositide 3-kinase angiogenesis (Fig. 1c, d, h–k, Supplementary Fig. 1)16. (PI3K) mediated downstream signaling activation1–11. In order to From these studies, we predicted that exogenously supplying regenerate PIP2 consumed during signaling, diacylglycerol (DAG) PI substrates to the vasculature would rescue the aberrant EC must be recycled to CDP-diacyglycerol (CDP-DAG) through the phenotypes in Cds2-deficient zebrafish, bypassing the need for activity of two vertebrate CDP-diacyglycerol synthase enzymes, PI recycling. Indeed, delivery of a single “bolus” intravascular CDS1 and CDS21–8,12–17. CDP-diacylglycerol synthase (CDS) injection of any PI species downstream of Cds2 dependent activity is necessary to resynthesize phosphoinositide (PI), the recycling—PI, PI(4)P, PI(4,5)P2 or PIP3—reactivates EC activity base molecule required for the generation of downstream PI to largely reverse the angiogenic defects noted in 72 hpf cds2y54 derivatives, such as PIP2. Loss or knockdown of either one of the mutant zebrafish embryos (Supplementary Fig. 2A–M)16. two CDS genes in ECs in vitro or in zebrafish embryos in vivo To further confirm the role of PI recycling in angiogenesis, we results in reduced angiogenesis, while excess PIP2 promotes generated a zebrafish mutant disrupting inositol monophosphatase increased angiogenesis in wild type endothelium16. Importantly, (impa2y602) (Supplementary Fig. 3A). The impa2 gene encodes a the vascular defects observed in cds2y54 null mutant zebrafish critical enzyme in the phosphatidylinositol recycling pathway that embryos that lack Cds2 but retain Cds1 activity seem to occur in removes the final phosphate group from myoinositol before it is the absence of other obvious developmental abnormalities16, linked to CDP-DAG to regenerate PI (Fig. 1a, Supplementary suggesting that the endothelium is differentially more sensitive to Fig. 3B), a step critical for the resynthesis of all subsequent PI a reduced capacity to resynthesize phosphoinositides than other species. Like the cds2y54 mutants, zebrafish maternal impa2y602 cell types and tissues. mutants also display defects in trunk angiogenesis at 32 hpf Anti-angiogenic therapies have been of significant interest for a (Supplementary Fig. 1C–F), demonstrating that interfering with number of years due to their potential to combat tumor growth either of two different PI recycling enzymes (Cds or Impa) impairs through cutting the tumor off from the host blood supply18–24. developmental angiogenesis. Importantly, the vascular defects However, the ability of tumors to overproduce pro-angiogenic observed in cds2y54 and impa2y602 mutantsappearintheabsence ligands and overcome targeted therapies has hampered this of other obvious developmental abnormalities (see ref. 24 for a approach to date25,26. An alternative way to circumvent this discussion of zebrafish angiogenesis assays), suggesting that ECs are problem is to target the re-synthesis of critical signaling sub- more sensitive to partial reductions in phosphoinositide recycling strates, like phosphoinositides, that are consumed during intra- capacity compared to other cells and tissues. It is also worth noting cellular transduction of pro-angiogenic signals in ECs, thereby that for both cds2y54 and impa2y602 there are additional isoforms of harnessing the tumor’s own production of excess stimulatory each of these enzymes that remain functional, i.e. Cds1 and Impa1, ligands to deplete adjacent host ECs of the capacity to respond to generating animals that have compromised but not completely these signals1,13,14,16,27–29. absent PI recycling capacity. Here we show using zebrafish, human cells, and mice that ECs To further explore the connections between phosphoinositide deficient in phosphoinositide recycling are uniquely sensitive to recycling and pro-angiogenic signaling, we injected a CMV: increased stimulation by VEGFA and other angiogenic cytokines. vegfaa transgene35 driving ubiquitous expression of Vegfaa into Instead of promoting increased angiogenesis, VEGFA stimulation the cds2y54 mutant zebrafish. Instead of promoting increased of PI recycling-deficient endothelium suppresses angiogenesis vessel growth, Vegfaa overexpression in cds2y54 null mutants or and decreases pro-angiogenic signaling, suggesting highly sti- cds2 morpholino (MO) treated embryos results in dramatically mulated, actively angiogenic ECs might be differentially sensitive reduced angiogenesis—including in the trunk intersegmental to reduced PI recycling capacity compared to quiescent ECs. To vessels, central arteries in the brain, and sub-intestinal vascular examine whether the sensitivity of angiogenically active endo- plexus (Fig. 1c–g, Supplementary Fig. 1). Longitudinal imaging thelium to reduced PI recycling capacity could be harnessed as an beginning at 24 hpf reveals that vessels fail to sprout and extend anti-angiogenic anti-tumor approach, we suppressed PI recycling properly (Supplementary Fig. 4). Sensitivity to Vegfaa is highly in mice and examined the effects on growth of allografted tumors. specifictocds2y54 mutants, as other zebrafish vascular mutants Our results show that either systemic or EC specific suppression including plcg1y10, flk1y17,andetsrpy11 do not show reduced of PI recycling results in reduced tumor growth, reduced pro- angiogenesis when injected with CMV:vegfaa transgene (Supple- angiogenic signaling in the endothelium, and reduced tumor mentary Fig. 5). angiogenesis. Together, these findings suggest that inhibition of Human ECs in vitro also exhibit this seemingly paradoxical phosphoinositide recycling may provide a useful anti-angiogenic sensitivity to VEGFA stimulation when CDS2 activity is reduced. approach, and highlights the general potential of targeting re- Two independent siRNA targets were validated in human synthesis of rate limiting signaling substrates as a therapeutic umbilical vein endothelial cells (HUVECs), confirming that both strategy. suppressed CDS2 protein levels, disrupted HUVEC invasive capacity in vitro, and suppressed p-ERK1/2 and p-AKT signaling downstream of VEGFA stimulation (Supplementary Fig. 6A–C). Results Further, CDS2 siRNA treatment reduces HUVEC proliferative Inhibiting phosphoinositide recycling sensitizes ECs to pro- capacity but does not induce apoptosis (Supplementary Fig. 6D, angiogenic stimuli. Previously, we reported the discovery of E). Exposing CDS2-deficient HUVECs to increasing levels of a zebrafish mutant in the CDP-diacylglycerol synthase (cds2) VEGFA stimulation results in decreased migratory capacity gene with defects in angiogenesis16. CDS activity is required for in 3D endothelial invasion assays24 as compared to controls resynthesis of phosphoinositides (PI), including phosphatidyli- (Fig. 1h–l), similar to the reduced angiogenesis noted upon nositol-(4,5)-bisphosphate, or PIP213,14,16,30 (Fig. 1a). In ECs, VEGFA stimulation in Cds2-deficient zebrafish.

2 NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z ARTICLE

Role of cds2 in phosphoinositol recycling and VEGFR2 signal activation Excess VEGF suppresses angiogenesis in cds2 y54 mutants in vivo

ab gn = 50 n = 50 n = 50 n = 50 P VEGFA y54 100 P Wild type cds2 mutant P PI PI(4)P PIP2 VEGFR2 PIP2 c d 80 ISV sprouts stall P To DLAV P DAG PIP3 +IP3 60 To myoseptum IMPase P Dorsal aorta CDP-DAG P Control No sprouts PI(4,5)P2 PLCγ1 PI3K 40 P P CDS2 IP3 CDS1 P % Intersegmental e f vessel phenotype 20 PA DAG ISV fail to ERK1/2 AKT vegfaa 0 sprout Wild cds2y54 Wildcds2y54 Myo-inositol

Arterial Cell CMV: type type P Phosphate specification, proliferation, CMV:vegfaa angiogenesis survival injected

Excess VEGFA suppresses angiogenesis in CDS2-Deficient HUVEC in vitro h i HUVEC k l 100 siControl 80 transfected siCDS2 70 Media with siRNA 80 60 50 Collagen 40 gel 60 VEGFA siControl 30 control (%) 20 40 siCDS2 to paired 10 18 h 37 °C Average # j Proportion of invading 0 of invading cells 20 10 ng/ml 40 ng/ml200 ng/ml

0 VEGFA concentration 10 ng/ml 40 ng/ml 200 ng/ml Plane of VEGFA concentration imaging siCDS2

m Wild type n cds2 y54 mutant o cds2 y54 Mutant + High VEGF-A

VEGF VEGF VEGF

VEGFR2 CDS1 VEGFR2 CDS1 VEGFR2 CDS1 CDS2CDDDS2DSS2 CDCDS2DDS2S2S PIP2 CDS2 PIP2 PIP2

IP3 + DAG IP3+ IP3+ DAG DAG EC Some EC Little/No activation activation EC activation

p q DLAV r VEGFA reduces ERK1/2 activation in Excess VEGFA reduces PIP2 levels Cds2 mutant EC in vivo in CDS2-deficient HUVEC ISVs 160 120 rhVEGFA 140 (ng/ml) DA 120 100 100 Control siRNA 0 p-ERK1/2+ Immunostain 80 40 60 80 200 40

Image mid-sagittal cross (norm’d to control) 20 siRNA 60 0 CDS2 section through dorsal aorta 0 40 Relative phospho-ERK1/2 VEGF y54 + VEGF 200 Control y54

Relative PI (4,5) P level (%) 40 Measure p-ERK1/2 y54 0481216 cds2 – – + + fluorescence in endothelial vegfaa DNA – + – + Hours post VEGFA addition cell nuclei

We further examined whether zebrafish vessels and HUVECs significant effects, nor did it cause additive effects when PI4K2A show increased sensitivity to VEGFA when inhibiting PI4K2A, and PI4K2B were inhibited together, suggesting PIPK2B is the PI4K2B, and PIP5K1C—the kinases that directly phosphorylate PI more critical of the two isoforms in this context. to generate PI(4,5)P2 (Supplementary Fig. 7)9–11. As with cds2y54 Together, these data show that limiting PI availability via mutants, injecting exogenous CMV:vegfaa in zebrafish treated suppression of CDS2 or downstream kinases in the PI recycling with PI4K inhibitor exacerbates the angiogenesis defects noted pathway leads to anti-angiogenic effects that are exacerbated, not (Supplementary Fig. 7A–G). VEGF-promoted anti-angiogenic ameliorated, by increased stimulation with VEGFA. phenotypes are also recapitulated in HUVECs in which either PI4K2B or PIP5K1C are suppressed using siRNA (Supplementary Fig. 7H, I). Suppression of PI4K2B or PIP5K1C sensitizes HUVEC VEGFA stimulation results in reduced signaling in CDS2-deficient to exogenously supplied VEGFA, with in vitro angiogenesis ECs. Defects in CDS2 reduces the capacity of ECs to regenerate reduced to a greater extent as VEGF concentration is increased PIs for signal transduction (Fig. 1a, b, m, n)1,4–6,8,13,16,36,37. (Supplementary Fig. 7H, I). Suppression of PI4K2A did not cause We hypothesized that while initial and/or low-level VEGFA

NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z

Fig. 1 CDS2 dependent angiogenic sprouting defects in vitro and in vivo are exacerbated by exogenous VEGFA addition. a Schematic diagram of phosphoinositide recycling. CDS1, CDS2, and IMP enzymes facilitate regeneration of phosphoinositol after consumption of PIP2 (see Fig. S12 for details). b VEGFR2 signaling schematic (modified from refs. 36,60). c–g Confocal images (c–f) and quantitation (g) of trunk intersegmental vessels (ISV) in 32hpf Tg (fli1a:egfp)y1 WT siblings (c, e)orcds2y54 mutants (d, f) injected with control (c, d) or CMV:vegfaa (e, f) DNA. Bars in g measure ISV that have not sprouted (yellow), grown halfway up the trunk (blue), or formed a complete ISV (gray). Data is representative of three different experiments, n = 50 ISVs per treatment group. h–j HUVEC 3D invasion assay used to model angiogenesis in vitro (h), with representative images from control and CDS2 siRNA-treated cultures (i, j). k Quantification of HUVEC cellular invasion into collagen gels at VEGFA doses indicated (n = 4 collagen gels; data is representative of three independent experiments). l Quantification of CDS2 siRNA HUVEC cellular invasion normalized to VEGFA dose-matched controls (see methods); Star indicates significance from control; plus indicates significance from individual VEGFA doses (t-test). m–o Schematic model for phosphoinositide recycling, CDS2, and angiogenesis. m Under normal conditions phosphoinositide recycling maintains PIP2 levels and VEGFA signal transduction. n Endothelium defective for CDS2 has a reduced capacity to recycle phosphoinositides, but under conditions of initial or low-level VEGFA stimulation sufficient PIP2 is regenerated to maintain VEGFA signal transduction. o During sustained and/or high-level VEGFA stimulation, however, PIP2 levels cannot be maintained, leading to a collapse of VEGFA signal transduction. p ELISA quantitation of PIP2 levels in control or CDS2 siRNA-treated HUVECs incubated with 0, 40, or 200 ng/ml VEGFA over a 16 h time course, normalized to levels in initial control siRNA-treated HUVEC without added VEGFA. Nine technical replicates were measured per sample, per experiment. Data is graphed as the average of two experimental replicates. q Diagram illustrating procedure for measurement of phospho-ERK1/2 in trunk endothelial nuclei by immunofluorescence. r Quantitation of trunk endothelial phospho-ERK1/2 in 30 hpf cds2y54 mutants and WT siblings ± (with/without) CMV:vegfaa DNA (column 1 n = 5 biologically independent animals; column 2 n = 4 biologically independent animals; column 3 n = 6 biologically independent animals; column 4 n = 7 biologically independent animals), data is representative of two independent experiments. Star indicates significance from control; plus indicates significance from cds2y54 mutant—CMV:vegfaa DNA condition (t-test). Bars = 100 μm. Box plots are graphed showing the median versus the first and third quartiles of the data (the middle, top, and bottom lines of the box, respectively). The whiskers demonstrate the spread of data within 1.5x above and below the interquartile range. All data points are shown as individual dots, with outliers shown above or below the whiskers.

stimulation of CDS2-deficient ECs might not substantially affect Fig. 9A–C). Furthermore, overexpressing either phospholipase C PI recycling (Fig. 1m, n), higher levels and/or sustained VEGFA gamma (plcγ1) or a combination of phosphoinositol-3-kinase stimulation would gradually deplete levels of PIP2 and other (pi3k) isoforms (both consume PIP2 during pro-angiogenic phosphoinositides, halting signal transduction (Fig. 1o). To signaling; see Fig. 1b) in zebrafish cds2y54 mutants leads to examine this directly we measured endogenous PIP2 levels over increased inhibition of angiogenesis (Supplementary Fig. 9D–J), time in control and CDS2-deficient HUVECs in vitro, challenged suggesting that inhibiting PI recycling capacity is a generalizable with VEGFA (Fig. 1p). In control siRNA-treated HUVECs way to interfere with Plcγ- or Pi3k-dependent tyrosine kinase VEGFA stimulation itself does not alter endogenous PIP2 levels mediated signaling pathways and is not solely restricted to over a 16-hour time course (Fig. 1p). PIP2 levels are also not blocking Vegfaa signaling. significantly reduced over 16 hours in CDS2 siRNA-treated HUVECs in the absence of exogenously added VEGFA, although the baseline level of PIP2 is slightly lower than in Inhibiting phosphoinositide recycling decreases tumor growth controls. In contrast, however, exogenous VEGFA stimulation of and angiogenesis. Based on our findings, we hypothesized that CDS2 siRNA-treated HUVECs results in a time-dependent phosphoinositide recycling might provide an effective target for decrease in PIP2 levels, with more rapid reduction in PIP2 anti-angiogenic anti-tumor treatment, as increased VEGFA and observed at higher VEGFA concentrations (Fig. 1p). Assessing other pro-angiogenic ligands secreted by tumors in response to PI(4)P, PIP2 and PIP3 levels side-by-side at 12 hours after vascular insufficiency and hypoxia might facilitate anti-angiogenic VEGFA stimulation, in lipid fractions collected from the same effects, rather than help to overcome them (Fig. 1o, see summary starting control versus CDS2-deficient HUVEC populations, schematic below). We used two different murine tumor allograft reveals that all three PI species are depleted in a VEGFA dose models, Lewis Lung Carcinoma (LLC)38,39 and B16-F10 (B16) dependent manner in roughly the same proportions (Supple- melanoma39, to determine whether systemic or EC specific sup- mentary Fig. 2N–P). Together, these data show that reducing the pression of phosphoinositide recycling could specifically inhibit phosphoinositide recycling by suppressing CDS2 limits the tumor growth and tumor angiogenesis while maintaining the availability of all downstream PI species. normal, healthy vasculature throughout the rest of the animal. We VEGF signaling downstream from VEGFR2 and PLCγ1is employed five different experimental paradigms to target phos- transduced via activation (phosphorylation) of ERK1/2 (Fig. 1b). phoinositide recycling in these tumor models: direct targeting of Injection of a CMV:vegfaa transgene into control zebrafish Cds2 using separate translation or splice blocking vivoMorpholi- embryos results in a modest increase in phospho-ERK1/2 levels as nos (vMO)40, inhibition of inositol monophosphatase (IMPase) assessed by western blot analysis of excised zebrafish trunk activity by the potent chemical inhibitor L-690,48841, inhibition of tissue (Supplementary Fig. 8), or a more marked increase by IMPase activity by lithium chloride (LiCl) treatment, and indu- immunostaining analysis examining phospho-ERK1/2 specifically cible, endothelial-specific genetic deletion of Cds2 in mice via the in trunk ECs (Fig. 1q, r). Cds2-deficient animals show somewhat Cre/Lox system. decreased phospho-ERK1/2, but injection of CMV:vegfaa trans- We administered murine Cds2 translation-blocking (vMO #1) gene into Cds2-deficient zebrafish results in a more dramatic or splice-blocking (vMO #2) vMO40, versus a control vMO, into decrease in phospho-ERK1/2, both by western blot (Supplemen- mice allografted with LLC (Fig. 2a) or B16-F10 (Supplementary tary Fig. 8) and by immunostaining analysis (Fig. 1r). Fig. 10A) tumors. The vMOs were introduced directly into the The seemingly paradoxical, negative effects of increased pro- circulation via tail vein or retro-orbital injection to increase angiogenic stimulation in Cds2-deficient animals are not vascular delivery and ensure their bioavailability, beginning restricted to Vegfaa. CDS2-deficient HUVECs treated with FGF two days prior to the implantation of tumor cells into each flank or EGF show similar reductions in signaling activation (p-AKT and continuing daily throughout the time course of tumor levels) to those seen with VEGFA treatment (Supplementary growth (12–18 days). LLC (Fig. 2b, c) or B16-F10 (Supplementary

4 NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z ARTICLE

a LLC tumor allograft treatment regimens

Daily IV injections Collection of samples for Representative of L-690,488 beginning immediately PIP2 analysis, IHC, and WB followingfollowing tumor injection into mice signaling studies LLC tumors

Day –5 Day –2 Day 0 Day 4 Day 8 Day 12–18 LLC tumor cells injected into each flank at Control vMO day 0 DailyDaily IV injections Tumor cells injected Start daily volume Take final tumor weight, of vMvMOO bebeginsgins two bilaterally into measurements once volume, and vessel density daysdays prior to tumor mouse flanks tumors become visible measurements injectioninjection into mice Cds2 vMO

Cds2 vMOs inhibit LLC tumor allograft growth L-690,488 inhibits LLC tumor allograft growth

bcdTumor volume Tumor final weight Tumor vascular density hijTumor volume Tumor final weight Tumor vascular density 250 300 Control (n = 8) 180 200 500 800 L-690,488 (n = 19; 80uM) Control vMO ) 250 200 3 400 140 ) Cds2 vMO #1 150 3 600 200 Cds2 vMO #2 150 300 100 100 150 100 400 200 60 Control 100 50 50 volume (mm 100 200 50 20 volume (mm LLC average tumor LLC average tumor LLC average tumor LLC average vessel LLC average vessel weight (% of control) weight (% of control) density (% of control) 0 density (% of control) 0 0 0 LLC average tumor 0 468101214 789101112 Days post-tumor injection Control Control Control Control Days post-tumor injection vMO #1 vMO #2 vMO #1 vMO #2 L-690,488 L-690,488 Cds2 Cds2 Cds2 Cds2 Cds2 vMO #1

PIP2 rescues Cds2 vMO inhibition of Myo-inositol rescues L-690,488 inhibition of LLC tumor growth LLC tumor growth e f g kl Tumor volume Tumor final weight Tumor vascular density Tumor volume Tumor final weight Control vMO Control vMO 1250 250 3500 180 Cds2 vMO #2 Cds2 vMO #2 Control Control 200 3000 1000 Myo-inositol 200 Cds2 vMO #2 140 2500 150 L-690,488 Cds2 vMO #2 + Carrier 750 150 2000 100 L-690,488 + myo-inositol Cds2 vMO #2 + PIP2 100 1500 500 100 60 1000 50 50 250 LLC average tumor weight (% of control) 500 LLC average tumor LLC average vessel

weight (% of control) 20 density (% of control) 0 0 Relative LLC tumor volume (% of starting control value) Relative LLC tumor volume (% of starting control value) 4 6 8 1012141618 468101214 PIP2 PIP2 Days post-tumor injection Carrier Carrier Days post-tumor injection Control Control Control L-690,488 (no PIP2) liposomes (no PIP2) liposomes Myo-inositol L-690,488 + myo-inositol (no liposomes) (no liposomes)

Fig. 2 Anti-PIP2 recycling therapies suppress tumor growth. a Schematic of Lewis Lung Carcinoma (LLC) tumor allograft assay with representative tumor images. Bar = 1 cm. b–g Cds2 vMO treatment (n = 10 biologically independent tumors for all groups). h–l L-690,488 small molecule inhibitor treatment (control n = 8 biologically independent tumors, L-690,488 n = 19 biologically independent tumors). Quantitation of average tumor volume (b, h), final tumor weight (c, i) and final tumor vascular density (d, j)at12–15 days post-tumor implantation, in control versus Cds2 vMO or L-690,488-treated animals (b–d versus h–j). d Quantitation of average tumor vascular density and representative images of CD31/PECAM labeled (green) versus DAPI (blue) LLC tumor sections from control vMO (top) and Cds2 vMO (bottom) treated mice. White arrowheads indicate sites of CD31/PECAM positive blood vessel labeling. Bar = 100 μm. For all vascular density measurement experiments: three images per tumor were acquired and vascular density measured for all groups. A minimum of two slide sections from each tumor in b–g (taken from sections at least 10 slices apart) were analyzed. Representative of three experimental replicates. e-g PIP2 “rescues” Cds2 vMO tumor inhibition. Quantitation of LLC average tumor volume (e), final tumor weight (f), and tumor vascular density (g) at 18 days post-tumor implantation in control vMO (n=8 biologically independent tumors), Cds2 vMO #2 (no liposomes, n = 9 biologically independent tumors), Cds2 vMO #2 + carrier liposome (no PIP2, n = 9 biologically independent tumors), or Cds2 vMO #2 + PIP2 loaded liposome-treated (n = 8 biologically independent tumors), LLC-allografted mice. Data in (e-g) are normalized to the average starting size of each individual tumor group at day 4, and shown as a percentage of the starting day 4 control (the PIP2 liposome injection start date). j Quantitation of average tumor vascular density of LLC tumor sections from control and L-690,488-treated mice. For all vascular density measurement experiments: three images per tumor were acquired and vascular density measured for all groups. A minimum of two slide sections from each tumor in h–l (taken from sections at least 10 slices apart) were analyzed. Representative of two experimental replicates. k, l Myo-inositol “rescues” L-690,488 tumor inhibition. Quantitation of LLC average tumor volume (k) and final tumor weight (l) at 15 days post-tumor implantation in control (untreated, n = 14 biologically independent tumors), myo-inositol (n = 11 biologically independent tumors), L-690,488 (n = 12 biologically independent tumors), or L-690,488 + myo-inositol (n = 10 biologically independent tumors) treated LLC-allografted mice. Data in (k, l) are normalized to the control condition. For all panels: p ≤ 0.05, error bars ± SEM. Star indicates significance from control (t-test). Box plots are graphed showing the median versus the first and third quartiles of the data (the middle, top, and bottom lines of the box, respectively). The whiskers demonstrate the spread of data within 1.5x above and below the interquartile range. All data points are shown as individual dots, with outliers shown above or below the whiskers.

Fig. 10B–D) allografted mice treated with either of the two Cds2 alter proliferation rates, suggesting the reduced tumor growth vMOs show decreased tumor volume and decreased ﬁnal tumor rates we note in our in vivo vMO studies are not the result of weight compared to control vMO-injected mice. To determine if inhibitory effects on the tumor cells themselves (Supplementary Cds2 vMO treatments have direct effects on tumor cell Figs. 10E and 11A). Quantitation of tumor vessel density in CD31 proliferation, we exposed isolated B16-F10 or LLC cells in vitro immunostained sections revealed ~60–80% reduction in vessel to the estimated comparable doses of Cds2 vMO that they would area in either LLC (Fig. 2D) or B16-F10 (Supplementary encounter in vivo and measured proliferation rates. In vitro Fig. 10F–I) tumor allografts from Cds2 vMO-treated animals exposure of LLC or B16-F10 tumor cells to Cds2 vMO does not compared to controls. Treatment with the vMOs does not change

NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z the overall mass of the mice (Supplementary Fig. 11B), and liver proliferation rates (Supplementary Fig. 13E, I). Exposure of LLC tissue from Cds2 vMO-treated mice showed no increase in or B16-F10 tumor cells to LiCl does not alter proliferation rates at Caspase 3-positive apoptotic cells or decrease in vascularization 10 and 20 uM doses, suggesting the reduced tumor growth rates compared to livers from control vMO-treated animals (Supple- we note in our study following LiCl treatment are not a result of mentary Fig. 11C–E). inhibitory effects on the tumor cells themselves (as noted above To further address the specificity of the Cds2 vMO treatments, and in Supplementary Figs. 10E and 11A for the Cds2 vMOs). we carried out “rescue” experiments using daily intravascular All of the treatments described above involve systemic inhibition administration of BODIPY-PIP2 liposomes, supplied in conjunc- of PI recycling. In order to show that the effects on tumor growth tion with the retro-orbital vMO injections in LLC tumor allograft are due to targeting of the tumor-associated host endothelium rather assays (Fig. 2e–g; Supplementary Fig. 11F, G; following a similar than the tumor cells themselves, we generated inducible EC-specific logic to that applied to the zebrafish studies in Supplementary Cds2 knockout (ECKO) mice by crossing the Cad5(PAC)-CreERT2 Fig. 2). Although there are limitations associated with exogen- EC specificCreline42 to Cds2tm1a(KOMP)Wtsi (i.e. Cds2lox/lox) ously supplying PI substrates, including difficulties delivering mice43. Tamoxifen was supplied to the indicated mouse groups for 5 lipids to desired cell types and subcellular locations, metabolism consecutive days prior to the implantation of LLC tumors into the of the lipids once they enter cells, and differing fatty acid flanks and then suspended for the remainder of the study. LLC acyl chain lengths on synthetic exogenous versus endogenous tumors from the Cad5(PAC)-CreERT2;Cds2lox/lox +TMX (homo- phosphatidylinositides, the goal of these experiments is to bypass zygous lox cassette, CreiECΔ, +TMX- experimental deletion group) the defect in PI recycling in CDS2-deficient animals by directly show decreased tumor volume and decreased final tumor weight providing a downstream PI substrate. Daily co-injection of PIP2 compared to all seven other control groups (Fig. 3a–d). Western blot liposomes into LLC-allografted mice treated with the Cds2 vMO data of EC’s isolated from LLC tumors from Cad5(PAC)-CreERT2; does reverse tumor growth and vascular density defects noted Cds2lox/lox +TMX mice shows maintained, significant suppression with Cds2 vMO treatment alone (Fig. 2e–g; Supplementary of CDS2 protein at the end of the 14 day experimental time Fig. 11F, G). Retro-orbital intravascular injections were chosen course (Fig. 3e). Examination of tumor vessel density by CD31 for these studies to make the liposomes more readily accessible to immunostaining of tumor versus liver sections revealed ~60% the vascular endothelium compared to surrounding tissues. reduction in tumor vessel area compared to controls, with no effects Indeed, immunostaining of tumor slide sections for the BODIPY on the liver vessel density (Fig. 3f–k).Importantly,forallinhibitor, fluorophore (to amplify signal), confirms enriched distribution of MO and ECKO treatments, routine daily veterinary checks of mice the PIP2 liposomes in the vasculature versus the surrounding in all groups did not reveal any adverse consequences, including in tumor tissue (Supplementary Fig. 11H). the ECKO group following tamoxifen treatment. Cds2-deleted A similar experimental paradigm was carried out using the ECKO mice ate normally and did not exhibit signs of illness potent inositolmonophosphatase (IMP) small molecule inhibi- throughout the course of the experiment. They also exhibited tor L-690,48841 in mice with LLC-allografted tumors. IMP normal social interaction, grooming, and other behaviors. Some of catalyzes removal of the phosphate group from inositol-1- the experimental groups implanted with tumors did become phosphate to generate myo-inositol, which is then combined moribund and exhibit some lethargy as tumors grew large near with CDP-DAG (the product of the reaction catalyzed by CDS) the point of final sacrifice. This was much less evident in the ECKO to regenerate phosphatidylinositol- PI (Fig. 1a). LLC tumors tumor-implanted mice, where tumor growth was restrained from mice treated with L-690,488 show decreased tumor volume compaired to the control tumor-implanted sibling groups which and final tumor weight compared to control DMSO-injected developed larger tumors prior to the point of sacrifice. These mice (Fig. 2h, i). Examination of tumor vessel density by CD31 findings, and similar observations from our vMO, L-690,488, and immunostaining of sections revealed ~50–60% reduction in LiCl experiments (Supplementary Figs. 11B, 12A, and 13N), suggest vessel area compared to controls (Fig. 2j). Treatment with that anti-PI recycling treatments, as performed in our studies, do L-690,488 does not affect the overall mass of the mice nor does notresultinharmfuleffectstothemice. it decrease normal vessel density in the liver (Supplementary The results described above show that anti-phosphoinositide Fig. 12). We carried out “rescue” experiments by co-injecting (PI) recycling treatments effectively reduce tumor growth via L-690,488 together with myo-inositol, the downstream product impaired tumor angiogenesis in LLC or B16-F10 allografted mice of the reaction catalyzed by IMP16,30,41,inordertobypassthe without significant effects on the endogenous vasculature of the reduced PI recycling capacity resulting from IMP inhibition liver in the same animals. We hypothesized that the increased (Fig. 2k, l). Myo-inositol administration effectively reverses sensitivity to anti-PI recycling treatments of tumor versus normal tumor growth inhibition caused by L-690,488, suggesting the vessels might reflect increased signaling and increased PI substrate effects of this drug on tumor growth are specific to inhibition of utilization in “activated” tumor-associated ECs compared to IMP activity (Fig. 2k, l). “quiescent” ECs in endogenous tissues and organs (Fig. 4a). We To further validate our L-690,488 results, we also targeted IMP examined this in two ways: (1) by ELISA quantitation of PIP2 activity by treating with LiCl, a compound already used in levels in ECs obtained from tumors and lungs collected at the humans. Lithium is an effective inhibitor of IMP, although it also termination of 14-day experiments from tumor-allografted has activity against other enzymes and pathways including WNT animals, and (2) by measuring active ERK1/2 and AKT in ECs signaling. We showed previously that LiCl inhibits angiogenesis from immunostained histological sections from tumors and livers in developing zebrafish in vivo and in EC invasion assays using of these same mice (Fig. 4b). Tumor ECs show substantially higher HUVEC in vitro, with the specificity of these effects for IMP phospho-ERK1/2 and phospho-AKT levels (Fig. 4c, i (upper left verified by myo-inositol rescue16. As with L-690,488, we find that panel), n (upper left panel)), as well as higher PIP2 levels (Fig. 4d) treatment of mice with LiCl inhibits tumor growth and tumor compared to liver or lung ECs (respectively). The increased angiogenesis in both LLC and B16-F10 allografts, and that these angiogenic signaling normally present in tumor-associated ECs, effects are largely reversed by co-administration of myo-inositol compared to ECs in endogenous host tissues, supports the idea (Supplementary Fig. 13). To determine if LiCl treatment has that systemic administration of PI recycling inhibitors within an direct effects on tumor cell proliferation, we exposed isolated appropriate therapeutic window could result in strong effects on B16-F10 or LLC cells in vitro to the estimated comparable doses “activated” tumor ECs with little or no effect on normal of LiCl that they would encounter in vivo, and measured “quiescent” ECs.

6 NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z ARTICLE

a LLC tumor allograft treatment regimen b

Daily IP injections Representative of tamoxifen for 5 days prior to LLC tumors tumor injection into Cds2lox mice

Day –5 Day 0Day 4 Day 8 Day 12–18 lox/lox LLC tumor cells Cad5(PAC)-CreERT2;Cds2 injected into each flank at day 0 Tumor cells injected Start daily volume Take final tumor weight, Cad5(PAC)-CreERT2) + TMX bilaterally into measurements once volume, and vessel density mouse flanks tumors become visible measurements

lox/lox Cad5(PAC)-CreERT2;Cds2 + TMX

Vascular specific Cds2 inactivation inhibits LLC tumor allograft growth

cdTumor volume Tumor final weight 300 2500 Wild type + TMX 250 200 2000 Cad5(PAC)-CreERT2) + TMX 150 Cad5(PAC)-CreERT2) 1500 100 lox/lox LLC average tumor

Cds2 + TMX weight (% of control) 50 lox/lox 1000 Cds2 0

lox/+ lox/lox Cad5(PAC)-CreERT2;Cds2 + TMX lox/lox ERT2) + TMX + TMX + TMX 500 Cre Cds2 lox/+ Relative LLC tumor volume lox/lox lox/lox lox/lox Cds2 (% of starting control value) Cad5(PAC)-CreERT2;Cds2 + TMX ERT2) + TMX Cds2 lox/lox Wild type + TMX (PAC)- Cds2 Cds2 ERT2; Cad5(PAC)-CreERT2;Cds2 Cre 0 Cre Cad5 ERT2; 4 5 6 7 8 9 10 11 12 13 14 (PAC)- ERT2; Cre Cre (PAC)- Days post-tumor injection Cad5 (PAC)- Cad5 (PAC)- Cad5 Cad5

Vascular specific Cds2 inactivation inhibits tumor induced angiogenesis CDS2 protein is suppressed in EC’s collected from conditional KO mice Representative tumor PECAM e f Tumor vascular density g Liver vascular density immunostaining images 250 200 h i 200 150 + TMX lox/lox lox/lox 150 100 Cds2 Cds2 100 Wild type Cad5(PAC)-CreERT2 ERT2; ERT2 +ERT2; TMX 50 + TMX + TMX Cre Cre Cre LLC average vessel 50 Liver average vessel density (% of control) density (% of control) (PAC)- (PAC)- (PAC)- j k 0 0 Cad5 Cad5 Cad5 CDS2 + TMX + TMX + TMX + TMX lox/+ lox/+ lox/lox lox/lox Tubulin ERT2 + TMX ERT2 + TMX Cad5(PAC)-CreERT2, Cad5(PAC)-CreERT2, Cds2 Wild type + TMXCre Cds2 lox/+ lox/lox Wild type + TMXCre Cds2 Cds2 Cds2 + TMX Cds2 + TMX ERT2; ERT2; (PAC)- ERT2; (PAC)- ERT2; PECAM (green) Cre Cre Cre Cre Cad5 Cad5 (PAC)- (PAC)- (PAC)- (PAC)- Cad5 Cad5 Cad5 Cad5

To determine whether systemic inhibition of PI recycling N). In contrast to the tumor EC findings, EC from lungs and preferentially reduces pro-angiogenic signaling in tumor- livers collected from the same LLC-allografted Cds2 vMO- associated ECs but not in quiescent endogenous ECs, we treated mice that the tumors were isolated from show no analyzed ECs from both tumors (Fig. 4e–i, o–q) and from livers/ reduction in levels of PIP2, p-AKT, or p-ERK1/2 (Fig. 4j-l, n; lungs (Fig. 4j–n, r–t) of the same Cds2 vMO (Fig. 4e–n) or L- Supplementary Fig. 14B–E).Importantly,tumorandliver/lung 690,488 (Fig. 4o–t) treated, LLC-allografted mice using immu- ECsfromthesameanimalsshowcomparablereductionsin nostaining and western blot analysis. Tumor-associated ECs CDS2 protein levels (Fig. 4h,m;SupplementaryFig.14A,F,G, from LLC-allografted, Cds2 vMO-treated mice show markedly L). Similar differential effects on tumor versus liver EC p-ERK1/ reduced levels of PIP2, p-AKT, and p-ERK1/2 (Fig. 4e–g, i; 2 and p-AKT activation were noted in mice treated with the Supplementary Fig. 14G–L). Consistent with previous reports in IMP inhibitor L-690,488 (Fig. 4o–t). the literature27–29,44,45 we find that p-AKT and p-ERK1/2 are Finally, to determine whether systemic inhibition of PI recycling more strongly activated in tumor-associated blood vessels than can slow the growth of pre-existing tumors, we allografted LLC in the surrounding tumor tissue, although longer-exposure tumors into mice and allowed the tumors to grow and establish. images confirm that p-AKT and p-ERK1/2 are indeed also Mice were then administered Cds2 translation-blocking (vMO #1) present in the surrounding tumor cells (Supplementary Fig. 14M, or splice-blocking (vMO #2) vMO, versus a control vMO daily

NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z

Fig. 3 Endothelial-specific genetic deletion of Cds2 promotes tumor growth inhibition. a Schematic of Lewis Lung Carcinoma (LLC) tumor allograft assay in endothelial-specific Cds2 knockout mice. b Representative tumor images from Vehicle Control (Cad5(PAC)-CreERT2;Cds2lox/lox); Cre, TMX Control (Cad5(PAC)-CreERT2 +TMX); and endothelial-specific Cds2 genetic deletion mice (Cad5(PAC)-CreERT2;Cds2lox/lox +TMX). Bar = 1cm. c, d Quantitation of LLC average tumor volume (c)andfinal tumor weight (d) at 14 days post- tumor implantation in the eight genetic conditions tested. Data in c are normalized to the starting day 4 tumor volume of Wild type + TMX controls. e Representative western blot images of EC protein isolated from tumors for each indicated genetic condition, probed for CDS2 and tubulin as a loading control. Blots demonstrate that CDS2 protein is still reduced in Cad5(PAC)-CreERT2;Cds2lox/lox mice at day 14, after only a 5 day pulse of TMX treatment initiated prior to the start of the experiment. p ≤ 0.05, error bars ± SEM. *Significance from control (t-test). f, g Quantitation of average tumor vascular density (f) and average liver vascular density (g)ofCD31/ PECAM labeled LLC tumor and liver sections from control Wild type +TMX, control Cad5(PAC)-CreERT2 +TMX, control Cad5(PAC)-CreERT2;Cds2lox/+ +TMX, and endothelial-specific Cds2 genetic deletion Cad5(PAC)-CreERT2;Cds2lox/lox +TMX mice. For all vascular density measurement experiments: three images per tumor were acquired and vascular density measured for all groups. A minimum of two slide sections from each tumor (taken from sections at least 10 slices apart) were analyzed. p ≤ 0.05, error bars ± SEM. Star indicates significance from control (t-test). h–k Representative images of CD31/PECAM labeled blood vessels in tumors from control Wild type +TMX, control Cad5(PAC)-CreERT2 +TMX, control Cad5(PAC)-CreERT2; Cds2lox/+ +TMX, and endothelial-specific Cds2 genetic deletion Cad5(PAC)-CreERT2;Cds2lox/lox +TMX mice. Bar =100 μm. Naming key: Wild type + TMX (no Cre, no Cds2 lox cassette; n = 18); Cad5(PAC)-CreERT2 +TMX (CreiECΔ only, +TMX; n = 13); Cad5(PAC)-CreERT2 (CreiECΔ only, no TMX vehicle control; n = 9); Cds2lox/lox +TMX (only homozygous Cds2 lox cassette, +TMX, no Cre; n = 6); Cds2lox/lox (only homozygous Cds2 lox cassette, no TMX vehicle control; n = 9); Cad5(PAC)-CreERT2;Cds2lox/+ +TMX (CreiECΔ, Cds2 heterozygous lox cassette, +TMX; n = 16); Cad5(PAC)-CreERT2; Cds2lox/lox +TMX (homozygous lox cassette, CreiECΔ, +TMX− experimental deletion group; n = 12); Cad5(PAC)-CreERT2;Cds2lox/lox (homozygous lox cassette, CreiECΔ, no TMX vehicle control; n = 7). Box plots are graphed showing the median versus the first and third quartiles of the data (the middle, top, and bottom lines of the box, respectively). The whiskers demonstrate the spread of data within 1.5x above and below the interquartile range. All data points are shown as individual dots, with outliers shown above or below the whiskers.

through retro-orbital injections (Fig. 5a). LLC-allografted mice We show that ECs are not only sensitive to reduced capacity to treated with either of the two Cds2 vMOs show decreased tumor re-synthesize phosphoinositides, but that this sensitivity is exacer- volume and decreased final tumor weight compared to control bated by increasing VEGFA stimulation (Fig. 1, Supplementary vMO-injected mice (Fig. 5b–d). Quantitation of tumor vessel Fig. 1). VEGFA has a well-documented role as a key pro-angiogenic density in CD31 immunostained sections revealed an ~70–80% ligand- increased VEGFA stimulation typically leads to increased reduction in vessel area in LLC tumor allografts from either of the angiogenic sprouting and growth of vessels. Our results show, Cds2 vMO-treated animals compared to controls (Fig. 5e–h). however, that VEGFA stimulation of ECs with compromised To determine whether systemic inhibition of PI recycling still capacity to recycle phosphoinositides has the seemingly paradoxical preferentially reduces signaling in tumor-associated ECs as opposite effect, resulting in less VEGFA-dependent signaling and compared to quiescent endogenous ECs in this model of pre- less vessel growth. This suggests that targeting phosphoinositide exisiting tumors, we analyzed ECs from lungs (Supplementary recycling might prove to be a useful approach for anti-angiogenic Fig. 15A–E) or tumors (Supplementary Fig. 15F–J) of the same cancer therapy, as tumors increasing VEGFA production in Cds2 vMO treated, LLC-allografted mice via Western blot analysis. response to oxygen or nutrient deprivation resulting from vascular ECs collected from the tumors have reduced PIP2, p-AKT, p- insufficiency caused by inhibition of tumor endothelial PI recycling ERK1/2, and CDS2 protein levels following Cds2 vMO treatment might actually increase, rather than decrease, the effectiveness of as compared to control treatments (Supplementary Fig. 15F–J), these types of inhibitors (Fig. 6).Wedemonstrateherethatinhi- while EC from lungs of the same Cds2 vMO-treated mice show biting phosphoinositide recycling in vivo does result in reduced minimal or no reduction in PIP2, p-AKT, or p-ERK1/2 levels tumor growth, reduced tumor angiogenesis, reduced angiogenic (Supplementary Fig. 15A–E), despite comparable reduction in signaling, and reduced PIP2 levels in the tumor-associated vascu- CDS2 protein levels. lature in multiple murine tumor allograft models without apparent Taken together, these data show that systemic inhibition of PI effects on the normal, quiescent vasculature of the same animals recycling causes angiogenic defects in the tumor-associated (Figs. 2–5 and Supplementary Figs. 10–15). vasculature, but has little effect on the pre-existing stable lung We would note again that CDS2 and IMP knockdown and the or liver vasculature of the same animals. other treatments we employ in this study only result in a partial loss of phosphoinositide recycling capacity. Our data suggests that in normal “resting” tissues, or even in moderately angio- Discussion genically active vessels, the residual capacity to recycle PIs by In this study we show that reduced capacity of ECs to recycle PI secondary CDS or IMP enzymes is adequate to regenerate suffi- substrates results in specific defects in angiogenesis and VEGF- cient PI substrates to maintain baseline intracellular signaling mediated signal transduction that correlate with reduced levels of (Fig. 4 and Supplementary Figs. 14, 15). On the other hand, under PIP2 in highly active EC populations, such as the tumor vascu- conditions of high angiogenic stimulation, such as in B16-F10 or lature. Zebrafish lacking CDP-diacyglycerol synthase 2 (Cds2), one LLC tumor-associated vessels or in highly angiogenically active of two enzymes required for re-synthesis of CDP-diacylglycerol developing vessels in the zebrafish, vascular defects appear (CDP-DAG) from diacylglycerol (DAG), show specificdefectsin (Figs. 1 and 4, and Supplementary Figs. 1, 3, 5). Although we did sprouting and growth of angiogenic vessels, while appearing not observe any obvious negative effects of our anti-PI recycling otherwise normal through early larval stages of development16. treatments in mice, it remains to be seen whether longer term Zebrafish with maternal depletion of the Impase, the enzyme anti-PI recycling treatments might eventually lead to some effects required for the re-sysnthesis of myo-inositol, show this same on angiogenically active normal tissues. However, some hints to defect. These results suggest that the highly active angiogenic ECs the long-term sustainability and effectiveness of anti-PI recycling present in developing vascular networks may be utilizing phos- therapies may possibly be gleaned from the historically well- tolerated and continued use of LiCl for management of bipolar phoinositide signaling and consuming PI substrates to a greater – extent than most other cell types. disorder and cognitive impairment in human patients30,46 48.

8 NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z ARTICLE

abNormal “Activated” tumor Tissue isolation from treated animals cdp-AKT and p-ERK1/2 PIP2 in tumor “quiescent” endothelium in tumor vs. liver ECs vs. lung ECs endothelium Lung 160 p-ERK1/2 120 Isolate EC For PIP2 p-AKT 140 100 measurement and 120 western blots 100 80 Tumor 80 60 Fix, embed 60 40 (% of tumor) (% of tumor) 40 Liver and section, Relative levels 20 norm’d to VECDN 20 Low signaling activation High signaling activation for IHC Relative PIP2 levels 0 0 Low PIP2 consumption High PIP2 consumption Tumor Liver umor Lung T Cds2 vMO treatment reduces PIP2, p-ERK1/2, and p-AKT levels in tumor EC from LLC allografted mice

ef g hi Representative tumor immunostaining images Tumor EC Tumor EC Tumor EC Tumor EC Control vMO Cds2 vMO #1 Cds2 vMO #2 PIP2 p-ERK1/2 p-AKT CDS2 120 180 160 140 160 140 100 120 140 120 80 100 120 100 100 p-ERK1/2 60 80 80 60 80 40 60 60 (% of control) (% of control)

40 (% of control) (% of control) 40 40 20 20 20 Relative p-AKT levels Relative PIP2 levels 20 Relative CDS2 levels Relative p-ERK1/2 levels

0 0 0 0 PECAM vMO: vMO: vMO: vMO:

Control Control Control Control Cds2 #1 Cds2 #2 Cds2 #1 Cds2 #2 Cds2 #1 Cds2 #2 Cds2 #1 Cds2 #2

Cds2 vMO treatment does not affect PIP2, p-Erk1/2, and p-Akt levels in lung or liver EC from LLC allografted mice

jklmn Representative liver immunostaining images Lung EC Liver EC Liver EC Liver EC Control vMO Cds2 vMO #1 Cds2 vMO #2 PIP2 p-ERK1/2 p-AKT CDS2 120 200 180 200 160 100 150 140 80 150 120 100 p-ERK1/2 60 100 100 80 40 60 Relative CDS2

(% of control) 50

50 Relative p-AKT 40

20 levels (% of control) levels (% of control) Relative PIP2 levels 20 Relative p-ERK1/2 levels (% of control)

0 0 0 0 PECAM vMO: vMO: vMO: vMO:

Control Control Control Control Cds2 #1 Cds2 #2 Cds2 #1 Cds2 #2 Cds2 #1 Cds2 #2 Cds2 #1 Cds2 #2

L-690,488 treatment reduces p-ERK1/2 and p-AKT L-690,488 treatment does not affect p-ERK1/2 and levels in tumor EC from LLC allografted mice p-AKT levels in liver EC from LLC allografted mice

opqTumor EC Tumor EC Representative tumor rstLiver EC Liver EC Representative liver p-ERK1/2 p-AKT Immunostaining images p-ERK1/2 p-AKT immunostaining images 200 200 200 200 Control L-690,488 Control L-690,488

150 150 150 150

100 100 100 100 p-ERK1/2 p-ERK1/2 (% of control) (% of control) (% of control) (% of control) 50 50 50 50 Relative p-AKT levels Relative p-AKT levels

0 Relative p-ERK1/2 levels Relative p-ERK1/2 levels 0 0 0 PECAM PECAM

Control Control Control Control L-690,488 L-690,488 L-690,488 L-690,488

Fig. 4 Systemic anti-PIP2 recycling therapies suppress VEGFR2 downstream signaling in LLC tumor models. a Model for signaling in activated versus quiescent endothelium. b Diagram illustrating tissues isolated from mice and their use. c Quantification of vascular phospho-ERK1/2 (black bars) and phospho-AKT (gray bars) levels in immunostained liver or tumor tissue from LLC-allografted control animals (n = 8). d ELISA measurement of PIP2 levels in lung or tumor endothelial cells from LLC-allografted control animals. Data is the average of three experimental replicates. e–n Cds2 vMO-treated, LLC- allografted mice. Quantification of PIP2 levels by ELISA in tumor (e) or lung (j) endothelial cells isolated from the same LLC-allografted animals. Data per treatment condition is graphed as the average of three experimental replicates. The ELISA measurements are normalized to whole cell tubulin lysates collected at the start of the lipid isolation procedure. Quantification of vascular phospho-ERK1/2 (f, k), phospho-AKT (g, l) and CDS2 (h, m) levels from immunostained sections of tumor (f–h) or liver (k–m) tissue collected from the same LLC-allografted animals. For all immunostaining quantitation experiments: three images per individual tumor or liver were acquired and signal intensity measured for all groups. A minimum of two slide sections from each tumor or liver (taken from sections at least 10 slices apart) were utilized for independent analysis. Tumors, n = 10; Livers, n = 5. i, n Representative images of tumor (i) or liver (n) sections from control or Cds2 vMO-treated mice, double immunostained with phospho-ERK1/2 antibody (green images) and PECAM antibody (red images). o–t L-690,488-treated, LLC-allografted mice. Quantification of vascular phospho-ERK1/2 (o, r) and phospho-AKT (p, s) levels from immunostained sections of tumor (o, p) or liver (r, s) tissue collected from the same LLC-allografted animals. q, t Representative images of tumor (q) or liver (t) sections from control (DMSO) or L-690,488-treated mice, double-immunostained with phospho-ERK1/2 antibody (green images) and PECAM antibody (red images). Control tumors, n = 8; L-690,488 tumors, n = 16; control livers, n = 4; L-690,488 livers, n = 8p≤ 0.05, error bars ± SEM. Star indicates significance from control (t-test). Bars = 200 μm. Box plots are graphed showing the median versus the first and third quartiles of the data (the middle, top, and bottom lines of the box, respectively). The whiskers demonstrate the spread of data within 1.5x above and below the interquartile range. All data points are shown as individual dots, with outliers shown above or below the whiskers.

NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z

a “Established” tumor allograft treatment regimen Daily IV injections Day 0 Day 4 Day 12–15 of vivoMorpholino Tumor cells beginning four days injected into after tumor injection each flank at into mice Tumor cells injected Tumors visible- Take final tumor weight, day 0 bilaterally into start daily injections of volume, and vessel density mouse flanks vivoMorpholinos and volume measurements measurements

Cds2 vMOs inhibit growth of established LLC allograft tumors bcd Tumor volume Tumor final weight 3000 140 Control vMO 120 2500 Cds2 vMO #1 Control 2000 Cds2 vMO #2 100 > 8) n > 6)

1500 n 80

60 Cds2

1000 vMO#1 volume (

weight ( 40 500 % Average LLC tumor LLC % average tumor 20 0 4567891011 12 13 14 0 Cds2

Ctrl Cds2 Cds2 vMO#2 Start of vMO Days post-tumor injection treatment vMO #1 vMO #2

Cds2 vMOs inhibit angiogenesis into established LLC allograft tumors e 160 fgh

120

40 LLC average vessel density (% of control)

0 Control Cds2 vMO #1 Cds2 vMO #2 Ctrl Cds2 Cds2 vMO #1 vMO #2

Fig. 5 Systemic Cds2 suppression inhibits growth of established tumors. a Schematic of established tumor allograft assay. Lewis Lung Carcinoma (LLC) tumor cells were injected into each flank of adult B6/C57 mice at day 0 and tumors allowed to develop for 12–15 days. Daily intravenous injections of control vivoMorpholino (vMO) or vMO targeting one of two independent sites in the Cds2 gene (vMO #1 and vMO #2) were started at day 4 after tumor implantation. Volume measurements were taken daily starting at day 4, when tumors became visible and vMO treatment started, with final tumor volumes and weights taken at the termination of the experiment. b Quantitation of daily average tumor volume (in mm3) of LLC tumors, normalized to the average starting size of each tumor condition at day 4 (the injection start date). c Quantitation of average tumor weight (in mg) of LLC tumors at 14 days post- tumor implantation. Control vMO (n = 9); Cds2 vMO #1 (n = 8); Cds2 vMO #2 (n = 6). d Images of tumors collected from control vMO or Cds2 vMO- treated animals at 14 days post-tumor implantation. Bar = 1 cm. e Quantitation of LLC tumor vessel density in control vMO- versus Cds2 vMO-treated mice. f–h Representative images of CD31/PECAM labeled (green) LLC tumor sections from control vMO (f) and Cds2 vMO (g, h) treated mice. White arrowheads indicate sites of CD31/PECAM positive blood vessel labeling. For all vascular density measurement experiments, three images per tumor were acquired (see tumor numbers above) and vascular density measured for all groups. A minimum of two slide sections from each tumor (taken from sections at least 10 slices apart) were utilized for independent analysis. Bar = 100 μm. Box plots are graphed showing the median versus the first and third quartiles of the data (the middle, top, and bottom lines of the box, respectively). The whiskers demonstrate the spread of data within 1.5x above and below the interquartile range. All data points are shown as individual dots, with outliers shown above or below the whiskers.

In our work, we demonstrate anti-angiogenic, anti-tumor incidence are inconclusive, likely due in part to small sample sizes effects of LiCl and L-690,488, two chemical inhibitors that target in the studies carried out and the complexity of the populations the IMPase enzyme to prevent PI/PIP2 regeneration (Figs. 2 and 4, being evaluated, with some studies reporting reduced rates of and Supplementary Fig. 13). Myo-inositol, the downstream pro- cancer mortality and morbidity49,50 and others not reporting such duct of IMPase that both chemicals block the production of30,is an association51–53. Additional, carefully designed studies using able to signiﬁcantly rescue the inhibitory effects elicited by these large patient populations will be needed to better examine whether drugs (Figs. 2 and 4, and Supplementary Fig. 13). This rescue there is a demonstrable link between cancer risk and lithium use. suggests that many of the anti-angiogenic, anti-tumorigenic effects However, as noted above, the historic use of lithium in patient we note in our inhibitor treatments are directly related to PI populations suggests that at therapeutically appropriate doses, recycling and not off-target effects, though the exact mechanism treatment with IMPase inhibitors can be well-tolerated over time. requires further investigation. Since lithium is already used in the In conclusion, our ﬁndings highlight the role that phosphoi- management of bipolar disorder30, it would be of interest to nositide recycling plays in tumor angiogenesis, and suggest that determine whether patients treated with lithium display a reduced targeted inhibition of PI recycling may provide a useful anti- risk of cancer or cancer related morbidity compared to the general angiogenic modality for the treatment of cancer (Fig. 6). The population. At present, the epidemiologic data related to cancer seemingly paradoxical sensitivity of ECs subjected to anti-PI

10 NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z ARTICLE

y54genotypeR- 5′-ATGAGGGTGTGGATGATGATGATA-3′.PCRproducts a Treatment with VEGF inhibitors were then digested with MseI cutting the PCR products carrying mutant cds2 into 193 and 124 bp fragments versus wild-type cds2 products that do not cut generating a 317 bp PCR product. All ﬁsh were housed as pairs in genotyping tanks (R&D Aquatics).

Zebrafish expression constructs, CRISPR and morpholinos. CMV:vegfaa- pCS2 VEGF VEGF VEGF + ( )-zvegf165 was a gift from Dr. Ruowen Ge (National University of Singapore). fi Relapse of vascular impairment CMV promoter-driven zebra sh vegf165 DNA for injection was generated as previously described. DNA was injected at a final concentration of 75 ng/μL. Zebrafish b Treatment with anti-PI recycling inhibitors pCR-plcg1 was MluI digested and pCSDest-pik3cd or pik3c2b was ClaI digested for 1 hour prior to DNA injection at the single cell stage at 75–100 ng/μL. Morpholino antisense oligonucleotides (Genetools) used in this study include: cds2 ATG MO: 5′-TCGCTGTCGTAATTCTGTCATGGTG-3′ that targets −4 to 21 of the 5′ untranslated region and coding region of cds2 (1.8 ng was used as a maximal dose); vegfaa:5′-CTCGTCTTATTTCCGTGACTGTTTT-3′ (6 ng was used as a VEGF VEGF VEGF maximal dose). The cds2 and vegfaa morpholinos used have both been previously 16,56 Decreased angiogenesis validated . Morpholino injections were performed on zebrafish embryos in the Decreased tumour growth 1– cell stage as previously described in ref. 2. Maximal doses were determined VEGF promotes vascular impairment by titration of morpholinos to levels that yielded maximal vascular-specific phenotypes with little to no nonspecific toxic effects. Fig. 6 Proposed model for inhibition of tumor growth with anti-VEGF PI4K inhibitor (PIK-93, #B-0306, Echelon Biosciences) was resuspended in versus anti-PI recycling therapies. a Currently available anti-angiogenic DMSO and used on the zebrafish at 20-40 uM. therapies targeting VEGF lead to an initial decrease in tumor growth and The impa2 y602 mutant allele was generated using the CRISPR/Cas9 system. The angiogenesis. However, production of high levels of pro-angiogenic ligands following guide RNAs were transcribed in vitro using the MEGAscript T7 Kit (Invitrogen), and injected at a dose of 150 pg/nl per embryo: by tumors results in a relapse in vascular impairment and return of tumor TAATACGACTCACTATAGGGCGTCAGGTTTATTGGTGGTTTTAGAGC growth. b Our results indicate that targeting PI recycling leads to reduced TAGAA. tumor vasculature and tumor growth, and further suggest that the effects pT3TS-nCas9 (Addgene) was transcribed using MEGAscript T7 Kit on tumor angiogenesis and tumor growth may be strengthened, not (Invitrogen), and injected at a dose of 300 pg/nl per embryo. Embryos were injected fi overcome, as tumor production of VEGFA and/or other cytokines at the single cell stage, screened for cutting ef ciency and grown on system. F1 generations were analyzed for mutations, and pairs crossed for analysis in the F2 increases. and beyond generations. ABI 3130xl Fragment Analyzer Protocol for Genotyping impa2 y602 mutants. PCR protocol with AmpliTaq Gold DNA Polymerase 1×(10 μl) Rxn: 1 μl 10x μ μ μ recycling treatments suggests that high levels of tumor-secreted, PCR Gold Buffer; 0.5 l MgCl2 25 mM; 1 l 0.5 mM ABI Fwd primer; 1 l1mM pro-angiogenic cytokines may facilitate the effectiveness of anti- ABI Rev primer; 0.2 μl 10 mM FAM-M13 primer; 0.1 μl dNTP Master Mix; 0.1 μl PI recycling therapies rather than overcome them (Fig. 6). Our TaqGold polymerase; 1 μl of 1:10 diluted crude gDNA; 5.1ul H2O. data also raises the possibility that anti-PI recycling therapy might TaqGold PCR Program: 95 °C 10 min; 95 °C 30 s; 58 °C 30 s; 72 °C 30 s (1 min/kb); GoTo Step 2 x34; 72 °C 10 min; 15 °C Hold; Run on ABI immediately or store at 4 °C not only inhibit tumor angiogenesis, but be even more effective in the dark for 24 h max. against aggressive tumors overexpressing high levels of VEGFA ABI 3130xl Pate set-up: HiDi Formamide/ROX master mix- 0.2 μl ROX300HD; and other pro-angiogenic cytokines that have traditionally been 9.8 μl HiDi Formamide; add 10 μl of master mix to each ABI plate sample well; add μ fl the most refractory to treatment25,26. Additional studies will be 2 lof uorescent PCR product; cap wells and denature at 95 °C for 5 min; uncap all wells and replace with ABI plate septa to run on the 3130xl. Follow needed to determine whether anti-PI recycling does indeed pro- manufacturer directions to utilize the ABI 3130xl. vide a useful therapeutic modality for treatment of these aggres- Gene Specific primers for genotyping: sive cancers. impa2: FW-M13: TGTAAAACGACGGCCAGTCTTACTGTGACATGTTAATGTGA TG Methods RV-PIG Tail: GTGTCTTGCACAAAGTTACAGGTGCCGTCGATG Zebrafish methods. Zebrafish (Danio rerio) embryos were raised and maintained FAM-M13: 5′-/56-FAM/ TGTAAAACGACGGCCAGT-3′. as described54,55. The Tg(fli1a:EGFP)y1 transgenic zebrafish line was previously described2. The cds2y25 and cds2y54 mutants were identified in an F3 genetic screen in the Tg(fli1a:EGFP)y1 background as described previously16. The impa2 y602 siRNA transfection and validation. Invitrogen single SilencerSelect or Stealth mutant in the Tg(fli1a:eGFP)y1 background was newly generated for this manu- siRNAs for CDS2 versus a siRNA negative control (control) were purchased and μ script. Embryos imaged at developmental stages later than 36 hpf were treated with resuspended in H2O at a concentration of 10 M. siPORT Amine (ThermoFisher 1-phenyl-2-thiourea to inhibit pigment formation55. Zebrafish husbandry and #AM4503) was used as a transfection agent following manufacturer recommen- research protocols were reviewed and approved by the NICHD Animal Care and dation (siRNA target sequences in Supplemental Table 1). A double transfection Use Committee. protocol, with a day of “rest” between treatments, was used as previously described6,57 with a final concentration of 50 nM siRNA per transfection condition. Validation of CDS2 siRNA suppression was previously reported16 but reconfirmed cds2y25 cds2y54 y25 fi Genotyping assays for and mutants. cds2 : Zebra sh carrying here by western blot. The siRNA sequences utilized are as follows: y25 the cds2 mutation were genotyped using KBiosciences Competitive Allele- CDS2 #1 (Stealth): GAGUACAACAAUGACACCAACAGCU fi Speci c PCR genotyping system (KASP) assays. Assays were performed on zeb- CDS2 #2 (Silencer Select): GUUUUAUCAGAGGCCCUAAUU fi ′ fl ra sh adult or embryonic genomic DNA extracts. Primer mixes with two 5 uor- PIP5K1C (Silencer Select- Validated): GCGUCGUGGUCAUGAACAAtt fi ′ labeled oligos that bind to allele-speci c primers are provided with a common 3 PI4K2A (Silencer Select- Validated): CCAAAGAUAUCGGACCCUAtt ′ oligo, as per the manufactures design. y25FAM, 5 -CAGCGGGAGGAGCC PI4K2B (Silencer Select- Validated): GAUUGACCGUGCAAAAUCAtt. TCTTC-3′ and y25HEX- 5′-GCAGCGGGAGGAGCCTCTTT-3′; y25common, 5′- AAATGA AGCGATGGTATTTGCTGAGGATT-3′. Details on assay and primer design can be obtained from KBioscience. The following PCR program was Endothelial cell culture and assays. HUVECs (Lonza) were cultured in bovine designed to determine wild-type versus mutant PCR fragments: (1) 94 °C, 15 min; hypothalamus extract, 0.01% Heparin and 20% FBS in M199 base media (Gibco) (2) 94 °C, 20 s; (3) 61 °C 1 min; (4) GOTO step 2 9x and decrease temperature in on 1 mg/mL gelatin-coated tissue culture flasks. HUVECs were used from step 3 by 0.6 °C each cycle; (5) 94 °C, 10 s; (6) 55 °C, 1 min; (7) GOTO step 5 35x; passages 3–6. (8) 25 °C, 5 min; Plate Read and End. As primers are labeled with fluorophores, 3-dimensional (3D) in vitro angiogenesis assays were done in 2.5 mg/mL homozygous products express only one fluorescent signal while heterozygous collagen type I (BD Bioscience, Acid Extracted) gels, prepared including SCF (R&D products will result in a mixed fluorescent signal. Systems, #255-SC/CF), SDF1α (R&D Systems, #350-NS/CF) and IL-3 (R&D cds2y54: PCR was performed on zebrafish adult or embryonic genomic Systems, #203-IL/CF in the gel at 200 ng/mL and VEGFA (R&D Systems, #293-VE/ DNA extracts using REDTaq ReadyMix PCR Reaction Mix and the following CF) at the indicated doses described in the manuscript, essentially as described6,57. primers: y54genotypeF- 5′-AACAGCTTGATGTAGCACAGCAGAGTA-3′; HUVECs were seeded on the gel surface at a density of 40,000 cells/well.

NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z

Culture media included FGF (200 ng/mL; R&D Systems, #233-FB-025/CF), IGF-II secondary antibody in 5% Sheep Serum, 1% Roche Blocking Buffer in PBST; (7) (200 ng/mL; R&D Systems, #292-G2-250), and ascorbic acid (Sigma). Assays were wash with PBST and imaging analysis. For mouse tumor and tissue slide sections: fixed in 3% glutaraldehyde at the indicated time points and processed for future the same essential staining protocol was utilized following (1) incubation of the analysis. Cells were stained with 1% toluidine blue in 30% methanol to increase slides at 50 °C to melt the paraffin wax, (2) 3x, 10 min wash in Histoclear (xylene visualization before imaging. Data analysis was done by imaging the endothelial alternative), (3) 2x, 5 min wash 100% EtOH, (4) 5 min was, 90% EtOH, (5) 5 min cell invasive front at 3 depths below the monolayer of cells (~50 μM apart in wash 70% EtOH, (6) continue to above staining procedure. All tissues were depth). Images were obtained and the number of endothelial cells present at each mounted and sectioned using the services of Hisoserv, Inc, Germantown, depth counted; the counts at the 3 depths were added together to give a total Maryland. All slides stained for p-AKT and p-ERK1/2 were co-stained with number of invading cells per well. A minimum of five replicate wells was averaged PECAM/CD31 antibody to confirm vascular expression for quantification to give a mean ± SD within a single experiment. At least 2 experimental replicates purposes. were performed. Data is represented as a percentage of the 10 ng/ml VEGFA Quantification of immunostaining intensity was performed by ImageJ analysis Control siRNA condition (Fig. 1j- % Average number of invading cells) or as a software. Images were acquired using a Leica SP5 II confocal microscope. All percentage of invasion inhibition at each indicated VEGFA dose (Fig. 1h- images were acquired at the same intensity, step size and image resolution for calculated by taking % Invading number of CDS2 KD cells/% Invading number of analysis. Data is reported as the percent average intensity per region of interest size. Paired Control cells = % Invasion Inhibition). For all vascular density measurement and pERK1/2/pAKT fluorescent quantitation For signaling activation assays, endothelial cells were seeded as a confluent experiments three images per tumor were acquired (see tumor numbers in the monolayer for 24 h in full growth media. The morning of the assay, cells were figure legends) and average measurements taken for all groups. A minimum of two serum starved in basal 1x M199 media for 4 h. The indicated growth factors/ligands slide sections from each tumor (taken from sections at least 10 slices apart) were were then supplied at a dose of 40 ng/ml in the basal media for 10 min. Lysates utilized for analysis, data ± s.e.m. Vascular expression of p-AKT and p-ERK1/2 was were collected and analyzed following the methods outlined below. in every case confirmed by co-staining with PECAM/CD31 antibody.

PI isoforms quantification . siRNA-treated HUVECs were plated on collagen type Tumor allografts. All animal studies were carried out according to NIH-approved I-coated plastic wells, allowed to attach overnight and then treated with the indi- protocols, in compliance with the Guide for the Care and use of Laboratory cated doses of rhVEGFA for predetermined times prior to PI(4)P, PI(4,5)P and fi fi Animals. B16-F10 or LLC tumor cells were maintained in 10% FBS/DMEM growth PIP3 quanti cation. Quanti cation was done using a competitive ELISA Kits media and obtained from the JS Gutkind lab. Prior to tumor injection into B6/C57 purchased from Echelon Biosciences (Echelon Biosciences,#K-4000E, #K-4500, mice, the mice were shaved to remove hair at the site of tumor growth and facilitate #K-2500 s) or utilizing a direct ELISA protocol for measuring PIP2 levels as pre- fi fi 58 tumor identi cation and volume measurements. On the day of injection, the tumor viously modi ed and described . Lipids was extracted following Echelon Bios- cells were trypsinized and resuspended in PBS. In all, 200 μl of PBS/cell suspension ciences manufacturer recommendations. Absorbance was read using a plate reader containing 106 cells was injected into each flank. For LiCl studies, normal saline at 450 nm. No detergent was used for washes to maintain the PIP2 lipid integrity. (control), LiCl (400 mg/kg), myoinositol (400 mg/kg), or LiCl+myoinositol in Samples were assessed in triplicate, normalized to cell number as assessed by μ fi 200 L normal saline were injected daily IP. For vMO (Gene Tools) studies, control counting four individual high powered elds (trypsin treatment of the cells can vMO (sequence: CCCTGTCCCCTACTTCGCTCATGCT), Cds2 vMO #1 alter lipid levels and is not advised if it can be avoided), and results given as mean (CCTCTGCCG TAGTTCGGTCATGCT), Cds2 vMO #2 (GGCTTGTCAACGA relative PIP2 levels as compared to control untreated cells, error bars ± SD for a TACTTACAATCA) were injected through the tail vein or retro-orbitally at a dose single experiment or error bars ± SEM when averaging multiple experiments. At μ – of 40 l of 2000 nmol stock per mouse per day. vMO were stored at room tem- least 2 3 experimental replicates were performed. perature and heated at 65 °C for 15 min prior to use. To verify suppression of the For measurements of PIP2 levels in lung ECs and tumor-associated ECs, CDS2 gene in vMO-treated mice, endothelial cells were isolated from the lungs at competitive ELISAs protocols were done as described above. Results were the termination of the studies (see below), and lysed for protein analysis to verify normalized to input protein levels based on Western blot tubulin expression. protein suppression. L-690,488 chemical inhibitor was purchased from Tocris (#0682), resuspended in DMSO at 100 mM and supplied via retro-orbital injection fi Liposome injections in Zebrafish and mice. Purified PI(4,5)P2 liposomes at a daily dose of 80uM per animal daily. For endothelial-speci c genetic sup- (Echelon Biosciences, #P-9045; to note- the SN2 position is 6-carbons long; the pression of Cds2, tamoxifen was injected IP daily at a dose of 1 mg/mouse in corn SN1 position, where the BODIPY is attached, is 6-carbons long and joined with an oil for 5 days prior to allograft of the tumors. The Cds2 mouse strain used for this amide linker) were incubated in equal molar concentration (40 μM) with lipid research project was created from ES cell clone EPD0033_1_C08 obtained from KOMP Repository (www.komp.org) and generated by the Wellcome Trust Sanger carrier constructs (Histone H1 Carrier #2, Echelon Biosciences) for 10 min at room 43 temperature prior to intra-vascular injection into either the zebrafish or mice (via Institute, and genotyped following their explicit instructions . The Cad5(PAC)- retro-orbital injection). For zebrafish studies, a single 15 nl “bolus” injection of CreERT2 were used with permission from Dr. Ralf Adams, and genotyped fol- – lowing their experimental protocols42. Endothelial cells were isolated from both BODIPY labeled PIP2 was delivered over the course of 2 3 min into the sinus fi venosus at 48 hpf. Only fish showing substantial and predominantly vascular lungs and tumors at the end of the 14 day experiment to con rm Cds2 protein dispersion of the BODIPY labeled PIP2 were utilized for ISV quantification ana- suppression. Eight groups were generated: Wild type + TMX (no Cre, no floxed Cds2, tamoxifen added CONTROL) lysis. Data analysis was done at 72 hpf. This same protocol was followed for PI (#C- + 00M6, Echelon Biosciences), PI(4)P (#C-04M6, Echelon Biosciences), and PIP3 Cad5(PAC)-CreERT2 TMX (cre driver alone, tamoxifen added CONTROL) Cad5(PAC)-CreERT2 (cre driver alone, no tamoxifen CONTROL) (#C-39M6, Echelon Biosciences) liposome injections. For mouse PI(4,5)P injection lox/lox + fl studies, daily retro-orbital injections of BODIPY PIP2 liposomes were done Cds2 TMX (homozygous oxed Cds2, no cre, tamoxifen added μ CONTROL) (40 M) versus lipid carrier (Histone H1 Carrier #2, Echelon Biosciences) alone lox/lox fl conditions. Cds2 (homozygous oxed Cds2, no cre, no tamoxifen CONTROL) Cad5(PAC)-CreERT2;Cds2lox/+ +TMX (heterozygous EC-specific knockout) Cad5(PAC)-CreERT2;Cds2lox/lox +TMX (homozygous EC-specific knockout) Immunoanalysis and immunostaining. Immunoanalysis: Zebrafish trunk tissue Cad5(PAC)-CreERT2;Cds2lox/lox (cre driver, homozygous floxed Cds2,no was collected using no. In all, 55 superfine forceps to remove the head and dissect tamoxifen CONTROL). off the yolk ball. Tissue was directly lysed in 2x Laemmli Sample Buffer containing For all experiments, calipers were used to measure the approximate tumor 5% β-ME and a PhosSTOP tablet (Roche), 10 ul per embryo. HUVEC culture volume daily, with final weight measurements (mg) made at the termination of the lysates were harvested directly into the same lysis buffer described above. Anti- experiment. Tumors were allowed to grow until control tumors reached a volume bodies: p-ERK1/2 (#4370), T. ERK (#4695), p-AKT (#4060), T. AKT (#2920) (Cell of ~500–1000 mm3, day 18 was reached, or significant impairment to the mouse Signaling Technologies, all 1:1,000), Tubulin (Sigma; #T6199- 1:10,000), CDS2 motility from tumor burden was noted. (ProteinTech, #13175-1-AP-1:500), BODIPY (Invitrogen, #A5770- 1:1000) in 5% BSA. Secondary HRP-conjugated antibodies were purchased from Santa Cruz or Invitrogen and used at 1:2000 in 5% milk. Quantification of relative band intensity Measurement of tumor vascularization. B16-F10 or LLC tumors or livers from was performed by ImageJ image analysis software, with results shown from at least the treatment mice were collected for analysis and paraffin embedded for H&E and two independent HUVEC assays, two independent zebrafish clutches, or two-three immunofluorescent staining or preserved in OCT medium for cryo-sections. At independent western blots from pooled EC samples collected from a minimum of minimum, four livers or eight tumors of each type were embedded and serially 2-4 tumors/lungs. Data is reported as a percentage of control levels, and p-ERK1/2 sectioned for vessel density analysis. Immunofluorescent labeling of CD31/PECAM and p-AKT are normalized to total tubulin levels. (BD Biosciences #550274, 1:10) and active Caspase3 (Sigma, #C4748- 1:1000) was Immunostaining: Zebrafish and 3D collagen assays utilized for immunostaining done and slides counterstained with HOECHST (Molecular Probes #33342, analysis were fixed in 4% or 2% PFA, respectively, at 4 °C overnight. Both were 1:2000). Images were taken using a Leica Inverted TCS-SP5 II confocal microscope immunostained following the same basic protocol: (1) 30 min RT incubation in and 20x objective. Three images were taken randomly from within the liver or the Tris-Glycine; (2) 1 h RT incubation for permeabilization with 0.01% TritonX-100; tumor body and the average vessel area measured per high powered imaging field. (3) 2 h RT incubation in blocking solution (5% Sheep Serum, 1% Roche Blocking Average vessel area was measured using ImageJ, measuring the average CD31/ Buffer in PBST); (4) 1 h at RT or overnight 4 °C incubation with 1:1000 primary PECAM+ area per high power image field. The data are reported as a mean, error antibody; (5) wash with PBST; (6) 2–3 h RT incubation with 1:2000 AlexaFlour bars ± SEM.

12 NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z ARTICLE

Quantification of tumor cell proliferation. For tumor cell proliferation in the 6. Stratman, A. N., Davis, M. J. & Davis, G. E. VEGF and FGF prime vascular presence of LiCl: B16-F10 or LLC tumor cells were plated in 24-well plates in 500 μl tube morphogenesis and sprouting directed by hematopoietic stem cell of 10% FBS/DMEM. Cells were allowed to attach and LiCl or vMO added. Esti- cytokines. Blood 117, 3709–3719 (2011). mated concentrations of LiCl were determined proportionally based off of doses 7. Warren, C. M. & Iruela-Arispe, M. L. Signaling circuitry in vascular provided in vivo to the mice (whole mouse dose/approximate tissue volume = x morphogenesis. Curr. Opin. Hematol. 17, 213–218 (2010). concentration). In all, 10 μM (estimated equivalent concentration to mouse dose) 8. Weinstein, B. M. & Lawson, N. D. Arteries, veins, Notch, and VEGF. Cold and 20 μM (estimated double concentration to mouse dose) concentrations of LiCl Spring Harb. Symp. Quant. Biol. 67, 155–162 (2002). were added to the growth media daily and proliferation assessed over a multi-day 9. Chu, K. M., Minogue, S., Hsuan, J. J. & Waugh, M. G. Differential effects of the time course. Culture media was changed completely every 3 days for the duration phosphatidylinositol 4-kinases, PI4KIIalpha and PI4KIIIbeta, on Akt of the time course. For tumor cell proliferation in the presence of vMO: Con- activation and apoptosis. Cell Death Dis. 1, e106 (2010). centrations for the vMOs were determined based off of the estimated mouse blood 10. Graupera, M. et al. Angiogenesis selectively requires the p110alpha isoform of volume (~2 ml) and proportioned to the volume of tissue culture media present in PI3K to control endothelial cell migration. Nature 453, 662–666 (2008). μ the assay. A daily dose of 4 l of the 2000 nmol stock per well was added to 24-well 11. van den Bout, I. & Divecha, N. PIP5K-driven PtdIns(4,5)P2 synthesis: plates containing 500 μl of 10% FBS/DMEM. Media was changed completely every – fi regulation and cellular functions. J. Cell Sci. 122, 3837 3850 (2009). 3 days. For both assays, cells were xed using 2% PFA and stained with HOECHST 12. Halford, S., Dulai, K. S., Daw, S. C., Fitzgibbon, J. & Hunt, D. M. Isolation and dye. Five distinct images per well were assessed and the number of nuclei per well chromosomal localization of two human CDP-diacylglycerol synthase (CDS) counted. The data are reported as the averages between images ± SD. genes. Genomics 54, 140–144 (1998). 13. Im, E. & Kazlauskas, A. Regulating angiogenesis at the level of PtdIns-4,5-P2. Isolation of mouse lung and tumor-associated endothelial cells. Isolation of EMBO J. 25, 2075–2082 (2006). mouse lung or tumor microvascular endothelial cells was performed essentially 14. Im, E. & Kazlauskas, A. PtdIns-4,5-P2 as a potential therapeutic target for according to a previously described protocol59. Modifications are briefly described, pathologic angiogenesis. Expert Opin. Ther. Targets 11, 443–451 (2007). lungs were removed from the mice, washed in 10% FBS/DMEM, minced into 15. Inglis-Broadgate, S. L. et al. Isolation and characterization of murine Cds 1–2 mm squares and digested with Collagenase Type I (2 mg/ml, Gibco) at 37 °C (CDP-diacylglycerol synthase) 1 and 2. Gene 356,19–31 (2005). fi μ for 2 h with occasional agitation. The cellular digest was ltered through a 70 m 16. Pan, W. et al. CDP-diacylglycerol synthetase-controlled phosphoinositide cell strainer, centrifuged at 1.2 g and the cells immediately incubated with Dyna- availability limits VEGFA signaling and vascular morphogenesis. Blood 120, bead sheep anti-rat IgG (Invitrogen) coated with a rat anti-mouse ICAM-2 mAb 489–498 (2012). μ (3C4, 22.5 L of antibody per 150 ul of beads) at room temperature for 10 min. The 17. Volta, M. et al. Identification and characterization of CDS2, a mammalian bead-bound cells were recovered with a magnet, washed four times, and the cells homolog of the Drosophila CDP-diacylglycerol synthase gene. Genomics 55, immediately collected for protein lysates and PIP2 ELISA protocols. All isolation 68–77 (1999). steps were done in the presence of phosphatase inhibitors. 18. Folkman, J. Anti-angiogenesis: new concept for therapy of solid tumors. Ann. Surg. 175, 409–416 (1972). Statistics. Statistical analysis of data was done using Microsoft Excel or Plotly. 19. Bergers, G. & Benjamin, L. E. Tumorigenesis and the angiogenic switch. Nat. Statistical significance was set at a minimum of p ≤ 0.05, and is indicated in Rev. Cancer 3, 401–410 (2003). individual figure legends. Student t-tests, two tailed, were used when analyzing two 20. Carmeliet, P. Angiogenesis in life, disease and medicine. Nature 438, 932–936 groups within a single experiment. Box plots are graphed showing the median (2005). versus the first and third quartiles of the data (the middle, top, and bottom lines of 21. Carmeliet, P. & Jain, R. K. Angiogenesis in cancer and other diseases. Nature the box, respectively) using Plotly. The whiskers demonstrate the spread of data 407, 249–257 (2000). within 1.5x above and below the interquartile range. All data points are shown as 22. Kidd, K. R. & Weinstein, B. M. Fishing for novel angiogenic therapies. Br. J. individual dots, with outliers shown above or below the whiskers. Pharmacol. 140, 585–594 (2003). 23. Kerbel, R. S. Tumor angiogenesis. N. Engl. J. Med. 358, 2039–2049 (2008). Study approval. Zebrafish and mouse husbandry and research protocols were 24. Nowak-Sliwinska, P. et al. Consensus guidelines for the use and interpretation reviewed and approved by the NICHD and NIDCR Animal Care and Use Com- of angiogenesis assays. Angiogenesis 21, 425–532 (2018). mittee at the National Institutes of Health. All animal studies were carried out 25. Abdollahi, A. & Folkman, J. Evading tumor evasion: current concepts and according to NIH-approved protocols, in compliance with ethical regulations for perspectives of anti-angiogenic cancer therapy. Drug Resist. Updat. 13,16–28 animal testing and research outlined in the Guide for the Care and use of (2010). Laboratory Animals. 26. Bergers, G. & Hanahan, D. Modes of resistance to anti-angiogenic therapy. Nat. Rev. Cancer 8, 592–603 (2008). Reporting summary. Further information on research design is available in 27. Carlo-Stella, C. et al. Sorafenib inhibits lymphoma xenografts by targeting the Nature Research Reporting Summary linked to this article. MAPK/ERK and AKT pathways in tumor and vascular cells. PLoS ONE 8, e61603 (2013). 28. Phung, T. L. et al. Pathological angiogenesis is induced by sustained Akt Data availability signaling and inhibited by rapamycin. Cancer Cell 10, 159–170 (2006). The data sets generated and/or analyzed during the current study are available from the 29. Zhai, X. et al. PI3K/AKT/Afadin signaling pathway contributes to pathological corresponding author B.M.W. upon reasonable request. The source data underlying main vascularization in glioblastomas. Oncol. Lett. 15, 1893–1899 (2018). and supplementary figures are provided as a Source Data file. 30. Atack, J. R., Broughton, H. B. & Pollack, S. J. Inositol monophosphatase–a putative target for Li+ in the treatment of bipolar disorder. Trends Neurosci. – Received: 2 October 2018; Accepted: 8 February 2020; 18, 343 349 (1995). 31. Yu, P. et al. FGF-dependent metabolic control of vascular development. Nature 545, 224–228 (2017). 32. Oladipupo, S. S. et al. Endothelial cell FGF signaling is required for injury response but not for vascular homeostasis. Proc. Natl Acad. Sci. USA 111, 13379–13384 (2014). 33. Hirata, A. et al. Direct inhibition of EGF receptor activation in vascular References endothelial cells by gefitinib (‘Iressa’, ZD1839). Cancer Sci. 95, 614–618 1. Lawson, N. D., Mugford, J. W., Diamond, B. A. & Weinstein, B. M. (2004). phospholipase C gamma-1 is required downstream of vascular endothelial 34. Poole, T. J., Finkelstein, E. B. & Cox, C. M. The role of FGF and VEGF in growth factor during arterial development. Genes Dev. 17, 1346–1351 angioblast induction and migration during vascular development. Dev. Dyn. (2003). 220,1–17 (2001). 2. Lawson, N. D. & Weinstein, B. M. In vivo imaging of embryonic vascular 35. Liang, D. et al. Cloning and characterization of vascular endothelial growth development using transgenic zebrafish. Dev. Biol. 248, 307–318 (2002). factor (VEGF) from zebrafish, Danio rerio. Biochim. Biophys. Acta 1397, 3. Kamei, M., Isogai, S., Pan, W. & Weinstein, B. M. Imaging blood vessels in the 14–20 (1998). zebrafish. Methods Cell Biol. 100,27–54 (2010). 36. Hong, C. C., Peterson, Q. P., Hong, J. Y. & Peterson, R. T. Artery/vein 4. Ren, B. et al. ERK1/2-Akt1 crosstalk regulates arteriogenesis in mice and specification is governed by opposing phosphatidylinositol-3 kinase and MAP zebrafish. J. Clin. Invest 120, 1217–1228 (2010). kinase/ERK signaling. Curr. Biol. 16, 1366–1372 (2006). 5. Shibuya, M. & Claesson-Welsh, L. Signal transduction by VEGF receptors in 37. Liang, D. et al. The role of vascular endothelial growth factor (VEGF) in regulation of angiogenesis and lymphangiogenesis. Exp. Cell Res. 312, 549–560 vasculogenesis, angiogenesis, and hematopoiesis in zebrafish development. (2006). Mech. Dev. 108,29–43 (2001).

NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14956-z

38. O’Reilly, M. S., Pirie-Shepherd, S., Lane, W. S. & Folkman, J. Antiangiogenic 60. Hong, C. C., Kume, T. & Peterson, R. T. Role of crosstalk between activity of the cleaved conformation of the serpin antithrombin. Science 285, phosphatidylinositol 3-kinase and extracellular signal-regulated kinase/ 1926–1928 (1999). mitogen-activated protein kinase pathways in artery-vein specification. Circ. 39. Palumbo, J. S. et al. Fibrinogen is an important determinant of the metastatic Res. 103, 573–579 (2008). potential of circulating tumor cells. Blood 96, 3302–3309 (2000). 40. Morcos, P. A., Li, Y. & Jiang, S. Vivo-Morpholinos: a non-peptide transporter delivers Morpholinos into a wide array of mouse tissues. Biotechniques 45, Acknowledgements 613–614 (2008). 616, 618 passim. We would like to thank members of the Weinstein laboratory for their critical comments 41. Atack, J. R. et al. Effects of L-690,488, a prodrug of the bisphosphonate on the manuscript. This work was supported by the intramural program of the National inositol monophosphatase inhibitor L-690,330, on phosphatidylinositol cycle Institute of Child Health and Human Development, National Institutes of Health (NIH; markers. J. Pharmacol. Exp. Ther. 270,70–76 (1994). B.M.W.), the intramural program of the National Institute of Dental and Craniofacial 42. Wang, Y. et al. Ephrin-B2 controls VEGF-induced angiogenesis and Research, National Institutes of Health (NIH; J.S.G.), and K99/R00 Pathway to Inde- lymphangiogenesis. Nature 465, 483–486 (2010). pendence Award (4R00HL125683–02; A.N.S.). 43. Skarnes, W. C. et al. A conditional knockout resource for the genome-wide study of mouse gene function. Nature 474, 337–342 (2011). Author contributions 44. Kim, I. K. et al. Sox7 promotes high-grade glioma by increasing VEGFR2- Contribution: A.N.S, C.M.M., Z.W., O.M.F, M.F.M, M.C.B, V.N.P., S.A.P., D.C., A.E.D., mediated vascular abnormality. J. Exp. Med. 215, 963–983 (2018). J.J.Y, and T.M.K. performed experiments; A.N.S, C.M.M., Z.W., O.M.F, M.F.M, S.A.P, 45. Solorzano, C. C. et al. In vivo intracellular signaling as a marker of V.N.P., S.A.P., D.C., A.E.D., J.J.Y, T.M.K., G.E.D., J.S.G., and B.M.W. analyzed results antiangiogenic activity. Cancer Res. 61, 7048–7051 (2001). and made the figures; A.N.S, C.M.M., V.N.P., G.E.D., J.S.G., and B.M.W. designed the 46. Sun, A. Y., Woods, S., Findling, R. L. & Stepanova, E. Safety considerations in research and wrote the paper. the psychopharmacology of pediatric bipolar disorder. Expert Opin. Drug Saf. 18, 777–794 (2019). 47. Heard, D. S., Tuttle, C. S. L., Lautenschlager, N. T. & Maier, A. B. Repurposing Competing interests proteostasis-modifying drugs to prevent or treat age-related dementia: a The authors declare no competing interests. systematic review. Front Physiol. 9, 1520 (2018). 48. Duthie, A. et al. Recruitment, retainment, and biomarkers of response; a pilot Additional information trial of lithium in humans with mild cognitive impairment. Front Mol. Supplementary information is available for this paper at https://doi.org/10.1038/s41467- Neurosci. 12, 163 (2019). 020-14956-z. 49. Rosenshfir, E., Shemesh, P., Aizenberg, D. & Barak, Y. Reduced risk of cancer in patients with bipolar affective disorder. Neurosci. Med. 2,57–60 Correspondence (2011). and requests for materials should be addressed to B.M.W. 50. Cohen, Y., Chetrit, A., Cohen, Y., Sirota, P. & Modan, B. Cancer morbidity in Peer review information Nature Communications thanks Guangwei Du, Arjan Griffioen psychiatric patients: influence of lithium carbonate treatment. Med. Oncol. 15, and the other, anonymous, reviewer(s) for their contribution to the peer review of 32–36 (1998). 51. Crump, C., Sundquist, K., Winkleby, M. A. & Sundquist, J. Comorbidities and this work. mortality in bipolar disorder: a Swedish national cohort study. JAMA Reprints and permission information Psychiatry 70, 931–939 (2013). is available at http://www.nature.com/reprints 52. Lin, G. M. et al. Cancer incidence in patients with schizophrenia or bipolar Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in disorder: a nationwide population-based study in Taiwan, 1997-2009. fi Schizophr. Bull. 39, 407–416 (2013). published maps and institutional af liations. 53. Zaidan, M. et al. Increased risk of solid renal tumors in lithium-treated patients. Kidney Int. https://doi.org/10.1038/ki.2014.2 (2014). 54. Kimmel, C. B., Ballard, W. W., Kimmel, S. R., Ullmann, B. & Schilling, T. F. Open Access This article is licensed under a Creative Commons Stages of embryonic development of the zebrafish. Dev. Dyn. 203, 253–310 Attribution 4.0 International License, which permits use, sharing, (1995). adaptation, distribution and reproduction in any medium or format, as long as you give 55. Westerfield, M. The Zebrafish Book. (Eugene, University of Oregon Press, appropriate credit to the original author(s) and the source, provide a link to the Creative 1995) Commons license, and indicate if changes were made. The images or other third party 56. Herbert, S. P. et al. Arterial-venous segregation by selective cell sprouting: an material in this article are included in the article’s Creative Commons license, unless alternative mode of blood vessel formation. Science 326, 294–298 (2009). indicated otherwise in a credit line to the material. If material is not included in the 57. Koh, W., Stratman, A. N., Sacharidou, A. & Davis, G. E. In vitro three article’s Creative Commons license and your intended use is not permitted by statutory dimensional collagen matrix models of endothelial lumen formation during regulation or exceeds the permitted use, you will need to obtain permission directly from vasculogenesis and angiogenesis. Methods Enzymol. 443,83–101 (2008). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 58. Pan, W. et al. Wnt3a-mediated formation of phosphatidylinositol 4,5- licenses/by/4.0/. bisphosphate regulates LRP6 phosphorylation. Science 321, 1350–1353 (2008). 59. Choi, E. Y. et al. Del-1, an endogenous leukocyte-endothelial adhesion inhibitor, limits inflammatory cell recruitment. Science 322, 1101–1104 This is a U.S. government work and not under copyright protection in the U.S.; foreign (2008). copyright protection may apply 2020

14 NATURE COMMUNICATIONS | (2020) 11:1204 | https://doi.org/10.1038/s41467-020-14956-z | www.nature.com/naturecommunications